iXBT Labs - Computer Hardware in Detail

Platform

Video

Multimedia

Mobile

Other

ATI RADEON 9800 XT 256MB Card Review

October 5, 2003





Contents

  1. Video card features
  2. Testbed, test tools, 2D quality
  3. Quality issues: why this driver version is used
  4. Test results: Quake3 ARENA
  5. Test results: Serious Sam: The Second Encounter
  6. Test results: Return to Castle Wolfenstein (speed and quality)
  7. Test results: Code Creatures DEMO
  8. Test results: Unreal Tournament 2003 (speed and quality)
  9. Test results: Unreal II: The Awakening (speed and quality)
  10. Test results: RightMark 3D
  11. Test results: TRAOD: Benchmarking in paris5_4
  12. Test results: TRAOD: Benchmarking in paris1c
  13. Test results: TRAOD: Benchmarking in paris2g
  14. Conclusion



On September 30 ATI pompously announced two new products: RADEON 9800XT (former R3600) and RADEON 9600XT (former RV360). These two products now crown their own niches: High-End and Middle-End respectively.

Actually, they are nothing but overclocked versions of the previous chips and cards. The engineers worked well on improving the technological processes, and now we can see new revisions of the R350 and RV350 able to run at higher clock speeds. But the competition between NVIDIA and ATI is very tough, and the marketers decided to give new names to the new revisions :-).

But in spite of the new names they are still based on ATI RADEON 9800 PRO and NVIDIA GeForce FX 5900 Ultra. By the way, the overclocked version of the latter is codenamed NV38. As the RADEON 9800 is just a speedier variant of the old processor in the reviews below to see what the R300 and its successor R350 are able of.


Theoretical materials and reviews of video cards which concern functional properties of the GPU ATI RADEON 9500/9600/9700/9800

So, what's the difference between the R360 and its predecessor? Actually, only the clock speeds differ. The RADEON 9800 PRO has its core clocked at 380 MHz while the memory speed depends: the 128MB card has DDR memory working at 340 (680) MHz and the 256MB one has the 350 (700) MHz DDR-II memory.

The RADEON 9800 XT has the clock speeds of 412/365 (730) MHz and is equipped with the 256MB DDR. The memory modules are bigger, and they are 8 instead of 16.

Does such announcement make sense? As you know, the latest High-End  accelerators were pretty expensive, and the prices fall down too slowly. The recommended price for the newcomer is $499. I suppose that 256MB RADEON 9800 PRO will either leave the market (the production volumes of such cards weren't great), or step down the price ladder. The first is more likely. The RADEON 9800 XT will simply replace the odd 256MB version of the R350, which had expensive DDR-II memory of the unneeded 256MB size. Besides, it was $100 dearer. At the same time, the breakaway from the 128MB card was inconsiderable.

That is why the R360 was aimed at replacing unpopular and unsuccessful RADEON 9800 PRO 256MB, and strengthening ATI's leading position in 3D performance.

What about NVIDIA? Their response is the upcoming NV38. This is also actually an overclocked version of the NV35. And we will try to compare ATI's card with the GeForce FX 5900 Ultra overclocked up to the speeds of the NV38. But it's just a trial test because NVIDIA hasn't released it officially. The clock speeds of the NV38 are taken from the Anandtech's site.

Card

ATI RADEON 9800 XT 256MB




ATI RADEON 9800 XT 256MB
AGP x8/x4 interface, 256 MB DDR SDRAM memory in 8 chips on both PCB sides. It's the first time the 256Mbit chips are packed in BGA. Earlier 256 MB was built up from 8 chips in TSOP package or 16 BGA chips.

Hynix 2.5ns memory chips (marked HY5DU573222) correspond to 400 (800) MHz but the memory works at 365 (730) MHz. GPU clocked at 412 MHz. 256bit memory interface.






Comparison with the RADEON 9800 PRO, front view
ATI RADEON 9800 XT 256MB Reference card ATI RADEON 9800 PRO











Comparison with the RADEON 9800 PRO, back view
ATI RADEON 9800 XT 256MB Reference card ATI RADEON 9800 PRO











The design is slightly altered, though it's basically the same. The space left for the Rage Theater chip indicates that the card can support VIVO. The higher the memory speed the better cooling is needed. It's the first time when ATI uses heatsinks for the memory chips.

ATI RADEON 9800 XT 256MB

The cooler differs much from earlier used solutions. Previous cards had a small heatsink with a fan in the center, and this one is equipped with a massive copper heatsink with the fan shifted off from the core center.

A similar cooler design was first used one and a half years ago on a GeForce4 Ti based card. The principle of operation is that air gets into the closed sink and blows over the hot processor.

I'm glad ATI changes its attitude towards appearance of their products and now we can see a really beautiful cooler,which is also very effective. Besides, it's not big and the card takes only one AGP slot.

On the back you can see a copper plate that is pressed against the memory chips with two strainers (both sinks are fastened with the screws on top). A jut on the plate is pressed against the PCB part where the core is located.












Here's the graphics processor (like the R350, it has a mirror surface):

As I already mentioned, the cooler is pretty compact and low, that is why it doesn't lock the first PCI slot. The fan sits high but it's shifted well to the left, that is why installation of low-profile PCI cards won't hamper cooling.

Note that the card can support VIVO if ATI's partners install the Rage Theater chip on it.

Testbed and drivers

Testbed: 

  • Pentium 4 3200 MHz based computer:
    • Intel Pentium 4 3200 MHz CPU;
    • DFI LANParty Pro875 (i875P) mainboard; 
    • 1024 MB DDR SDRAM; 
    • Seagate Barracuda IV 40GB HDD; 
    • Windows XP SP1; DirectX 9.0a;
    • ViewSonic P810 (21") and ViewSonic P817 (21") monitors.
    • ATI drivers v6.378.

VSync off, S3TC off in applications. 

Cards used for comparison:

  • ATI RADEON 9800 PRO (380/350 (700) MHz, 256 MB DDR).
  • ASUS V9950 Ultra (GeForce Fx 5900 Ultra, 450/425 (850) MHz, 256 MB DDR, driver 45.23, 51.75).

Test results

Before we start examining 2D quality, I should say there are no complete techniques for objective 2D quality estimation because: 

  1. 2D quality much depends on certain samples for almost all modern 3D accelerators; 
  2. Besides videocards, 2D quality depends on monitors and cables; 
  3. Moreover, certain monitors might not work properly with certain video cards. 

With the ViewSonic P817 monitor and BNC Bargo cable the card showed excellent quality at the following resolutions and clock speeds:

ATI RADEON 9800 XT 256MB 1600x1200x85Hz, 1280x1024x120Hz, 1024x768x160Hz

Quality issues: why these driver versions are used

First of all, we do not carry out synthetic tests because it's obvious that overclocking won't change the situation much - the correlation of the test results will remain the same. You can look at the current situation from Alexander Medvedev's recent article.

Secondly, in comparison of the RADEON 9800 XT with its competitors we will use NVIDIA's drivers of the version 45.23 (official release, WHQL). The results obtained with the v51.75 will be also provided, but this version is still raw and has many flaws. Have a look at some examples from the RtCW and Unreal II.

RADEON 9800 FX 5900 driver 45.23 FX 5900 driver 51.75
Unreal II, Example 1









Unreal II, Example 2









Unreal II, Example 3











RADEON 9800/FX 5900 driver 45.23 FX 5900 driver 51.75
RtCW








The Unreal II game has problems with fog on the v51.75 while in the RtCW some objects are light-struck (the problems with multitexturing and lightmaps?). As you can see, the driver 51.75 speeds up the GeForce FX family exactly in these games. There are some other problems with quality on these drivers. Look for more information in our September's 3Digest.

That's why I can't use drivers of such quality in comparison with the performance on ATI's WHQL drivers. And in NVIDIA's case we used only v45.23 to compare the cards' performance.

The newer drivers will be used only in the TR:AoD benchmark where the Depth of Field functions incorrectly on v45.23 and doesn't show any bugs on v51.75. In general, this benchmark works more or less well on these drivers.

Test results: performance

Conventional signs: ANISO 8xP - Anisotropic 8x Performance (earlier it was called Balanced), ANISO 8xQ - Anisotropic 8x Quality, ANISO 16xQ - Anisotropic 16x Quality. 

Some time ago we decided not to compare ATI's maximum anisotropic quality of 16x to two NVIDIA's modes anymore. The ANISO 8x Quality mode provided the real maximum quality with both trilinear filtering and anisotropy working to their full capacity. The ATI 16x Quality showed sharper images due to the 16th degree but on some surfaces the filtering quality reduced. That's the way ATI's anisotropy works. That is why we thought it was more correct to compare exactly this ATI's mode with NVIDIA's Performance and Quality ones.

But NVIDIA's today's optimization policy changes the things and we do not know anymore if there are applications where NVIDIA's anisotropy works to its full capacity. That is why we consider it to be correct to compare ANISO 16xQ (ATI) to ANISO 8xQ (NV). Both have their strong and weak points, but in general they compensate each other.

Test applications: 

Return to Castle Wolfenstein (MultiPlayer) (id Software/Activision) - OpenGL, multitexturing, ixbt0703-demo, test settings - maximum, S3TC OFF, the configurations can be downloaded from here
 
Serious Sam: The Second Encounter v.1.05 (Croteam/GodGames) - OpenGL, multitexturing, ixbt0703 demo, test settings: quality, S3TC OFF
 
Quake3 Arena v.1.17 (id Software/Activision) - OpenGL, multitexturing, ixbt0703 demo, test settings - maximum: detailing level - High, texture detailing level - #4, S3TC OFF, smoothness of curves is much increased through variables r_subdivisions "1" and r_lodCurveError "30000" (at default r_lodCurveError is 250 !), the configurations can be downloaded from here
 
Unreal Tournament 2003 v.2225 (Digital Extreme/Epic Games) - Direct3D, Vertex Shaders, Hardware T&L, Dot3, cube texturing, default quality
 
Code Creatures Benchmark Pro (CodeCult) - the game that demonstrates card's operation in DirectX 8.1, Shaders, HW T&L.
 
Unreal II: The Awakening (Legend Ent./Epic Games) - Direct3D, Vertex Shaders, Hardware T&L, Dot3, cube texturing, default quality
 
RightMark 3D v.0.4 (one of the test scenes) - DirectX 8.1, Dot3, cube texturing, shadow buffers, vertex and pixel shaders (1.1, 1.4).

Tomb Raider: Angel of Darkness v.49 (Core Design/Eldos Software) - DirectX 9.0, three demo scenes, maximum quality, Depth of Fields PS20 off. The settings are equal for all the cards tested.









For the RADEON 9800 PRO and GeForce FX 5900 cards the game and the patch 49 were reinstalled after the cards from one camp were tested so that the game settings don't get mixed. If you need patch 49 which is not easy to find and the demo benchmarks let me know by email.

Quake3 Arena















The light modes without AA and anisotropy: there's no a definite leader - the 9800 XT and NV35 alternately take the crown. Emulation of the NV38 improved its performance a little.

AA enabled: the 9800 XT reigns except in 1024x768.

Anisotropy enabled: approximate parity.

AA & anisotropy enabled: RADEON 9800 XT wins.

Serious Sam: The Second Encounter















The light modes without AA and anisotropy: the cards look more or less equal.

AA enabled: the card loses the battle.

Anisotropy enabled: approximate parity.

AA & anisotropy enabled: RADEON 9800 XT falls a little behind its competitors.

Return to Castle Wolfenstein (Multiplayer)















The light modes without AA and anisotropy: the 9800 XT wins.

AA enabled: the same.

Anisotropy enabled: the victory is even brighter.

AA & anisotropy enabled: RADEON 9800 XT scores excellent results and shows a solid breakaway!

Code Creatures















The light modes without AA and anisotropy: the 9800 XT loses.

AA enabled: the 9800 XT loses even more (NVIDIA's optimizations work!).

Anisotropy enabled: the 9800 XT grasps the victory (NVIDIA's method filters almost every blade).

AA & anisotropy enabled: the RADEON 9800 XT is slightly ahead due to its anisotropy.


Unreal Tournament 2003















The light modes without AA and anisotropy: if it beats the 5900 Ultra it steps back in front of the NV38.

AA enabled: the 9800 XT wins.

Anisotropy enabled: the 9800 XT wins again, though the competition against the NV38 looks more complicated.

AA & anisotropy enabled: the RADEON 9800 XT wins!

Unreal II: The Awakening















In the light modes without AA and anisotropy and with these functions enabled the RADEON 9800 XT comes out victor!

RightMark 3D















The test is entirely based on shaders, which is a problem for the NV35/38. That is why the R9800 XT easily wins.

TR:AoD, Paris5_4 DEMO









This is the toughest test regarding shader speed.

No doubt ATI's card triumphs.

TR:AoD, Paris1c DEMO









This scene uses fewer shaders that is why the GeForce FX comes closer to the RADEONs, though the gap is still wide.

TR:AoD, Paris2g DEMO









This test has few shaders, and the cards perform very close to each other, though the RADEON is still ahead.

Conclusion

If we sum it up, the RADEON 9800 XT looks a new 3D king. There are some tests where it yields to its counterpart but remember that we used the overclocked NV35 as the NV38, i.e. NVIDIA's driver didn't know that we wanted the speed of the NV38: the DeviceiD usually has an effect on the driver optimizations. I hope the emulated NV38 won't differ much from the real one, and we won't have to retest the cards.




It's also obvious that the RADEON 9800 XT does not press down the former top card but replaces the unsuccessful 256MB RADEON 9800 PRO. That is why they have equal prices. But we don't know how NVIDIA  is going to position its NV38. Probably, their product will also replace the 5900 Ultra.

If you already have a card of the level of RADEON 9800 and GeForce FX 5900, it's not worth replacing them as the new-comers  are not attractive that much. The speed gain is not great.




Those who are ready to pay $500 and haven't made their choice should definitely consider the RADEON 9800 XT. It doesn't make sense to go with the RADEON 9800 PRO, and the GFFX 5900 Ultra is slower in shaders (a card for $500 is not going to be used for old games, right?). Although this is a marketing product, it will get a certain marketshare as thanks to its leadership. Stay tuned!

In our 3Digest you can find full comparison characteristics for video cards of this and other classes.



Andrey Vorobiev  (anvakams@ixbt.com)

Write a comment below. No registration needed!


Article navigation:



blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook


Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.